Researches at Stanford have developed a neural network, CheXNet, capable of outperforming Radiologists at spotting Pneumonia and 13 other diseases. Advances in the capabilities of AI has enabled neural networks to recently perform as good or better than doctors at spotting problems on CT scans, and this latest result adds spotting lung diseases to that list.
Trained using 100,000 chest x-rays that contained 14 different diseases from a publicly available data source, CheXNet was initially tested against four Radiologists to see who could better spot Pneumonia. Not only did CheXNet spot Pneumonia better, it also outperformed the Radiologists at spotting the 13 other dangerous diseases it was trained with. Pneumonia remains a worldwide threat, and remains the single largest infectious cause of death to children worldwide. Advancements in the early detection of this dangerous lung disease could greatly decrease it's impact and provides more support to the growing sentiment that AI should be relied upon more in the medicine. Deep learning algorithms excel at analyzing and learning from image-based data sets and research is beginning to show that these tasks represent an area where AI consistently outperforms humans.
Click here to read the full article.
A2D Digital Feed
Follow the leading stories about digital transformation.